Federal gosudarstvennoe uchebnoe predpriyatie Chair of the System of Artificial Intelligence

4
0
4 0.43
0
17
17 0.33
X23
3
0
X30 1.00
16
52
12
4 0.75
9 0.48
X11
9
52
17 1.00
17
0
13
0
1
17 0.39
8
17 0.25
TH
10
dnee
Rang
2 0.28
17 0.29
44
0
0
0
0
X28 0.75
7 0.19 48
0
0
SL
0
10
12
3
15 1.00
17 0.32 43
13
17 0.47
WN 0
3 0.88
1
5
0
52
ran
11
X20 0.62
1
Table 3.5. - Results opredeleniya znachimostey vxodnyx
1
0
0
Set
17 0.49
3 0.57
17
41
17
7
6
3
gov
Set
0
5 0.30 48
2
FR 0.97
4 0.17
0
17 0.45
11 0.23
0
0
5
STU 0.53
6 0.59 27
parameters after contrast (fragment for 4 of 10 sets)
2
17 0.27
12
X22
X6
7 0.75
2
9 0.37 34
ma
17
6
17 0.71
17 0.79
53
4
17 0.60
17 0.67
55
3
ST 0.79
0
Rang
14 0.37 36
9
17 0.64
17 1.00
5
4
3 0.53
8 0.26
9
17
0
6 0.46
0
14 0.32
TU
Rang
0
17 0.16
2 0.21
6 0.28
8
X4
Sum
17
6
0
4
1 0.36
KR
11
2
17 0.61
1
MN 0
Rang
14
51
14
0
0
17 0.58
57
17 0.30
0
Set
5
17
31
17
0
5
AU
17 0.25
4
15
17
0
1
8 0.34
53
17 0.46
0
0
0
3
LIT
WIN 0
17 0.30
2 0.30
0
0
50
Set
1
13 0.15
0
12 0.44
12
17 0.75
4
9 0.63
0
Sre
52
17 0.58
12
17 0.31
17 0.82
2
17 0.19
106
Machine Translated by Google

---

17 0.12 63
17 0.13
18
22
17
0
10 0.10
17 0.46
15
13
17 0.11
23
0
Frequency in practical research ispolzuyut dopolnitelnÿe
0
0
17
7
0
58
udovletvoritelnoe - the process repeats, if it is net - stop
contrast and binaryization.
17 0.53
17
0
0
21
11
62
contrast of input parameters) and znachimosti of input parameters. It is useful for comparison
and analysis of results, received before and after
ZAR 0
17
SU
0
17 0.48
vÿpolneniya analogichen opisaniyu metodiki kontrastirovaniya. Pri
66
seti - syndromes pervogo urovnya i odnovremenno vxodnymi symptomami
17
X2
0
0
17
vesa ne menyayutsya when povtornom obuchenii.
0
64
0
0
16
parameters, neobxodimyy dlya resheniya zadachi (dlya etogo ispolzuyut
20
0
cherez punkt «Verbalizatsiya» s soxraneniem vsex faylov), poluchennyx posle
17
17
binarizuetsya otdelnÿy synapse, set douchivaetsya. Esli kachestvo obucheniya
17 0.38
17 0.09
17 0.11
synapses (when soxranenii in the name of the project insert a letter b). Poryadok
yavlyayutsya vxodnymi symptomami, vyxodnÿe signalaly neyronov pervogo sloya
0
X19
17 0.13
0
68
0
17
soxranyaetsya, t. e. binarizovannÿe synapses ostayutsya binarizovannymi i ix
14
17
12
0
17 0.49
17 0.12
17
struktury neskolkix neyrosetey (tekstovoe opisanie mojno poluchit
0
0
17 0.45
AN
17
0
61
process proisxodit sxodnÿm kontrastirovaniyu obrazom: posledovatelno
X8
SVA 0
61
0
10
19
operatsii sovmestno. For example, opredelyayut minimum set vxodnyx
17 0.52
17
It is possible to test experiments on binary weights
0
65
Verbalization traktuetsya kak process vosstanovleniya symptom syndromnoy
struktury ponyatiy predmetnoy oblasti. Input signal sets
uproshcheniya procedure.
G3. Verbalization. Mojno privesti graphic representation
17
X18
0
17
algorithm. V xode doobucheniya struktura seti posle binarizatsii
17 0.00
for neurons vtorogo sloya, generiruyushchix syndromes vtorogo urovnya, i t. d.
binarizatsii (privedenii vesov sinapsov k zadannym znacheniyam) dannyy
0
X16 0.38
0
107
Machine Translated by Google

---

SVAR = (SVAR-0.5) / 0.5
Predobrabotka vxodnyx poley BD dlya podachi seti:
Syndrome 2-go level:
KRAN = (KRAN-0.5) / 0.5
Syndrome1_1 = Sigmoida1 (ELEC-0.16 * CRANE-0.31 * SVAR + 0.06 * X22)
Syndrome1_2 = Sigmoida1 (0.24 * X11) Syndrome1_3 = Sigmoida1 (0.36 *
FR-0.37 * X9- 0.10 * X10)
after contrast. Polya bazy
dannyx (exodus symptom):
X9 = (X9-10.1) / 10.1
Syndrome3_1 = Sigmoida3 (0.50 * Syndrome2_1 + 0.42 * Syndrome2_2)
FR = (FR-0.5) / 0.5
Privedem primer verbalnogo opisaniya dlya odnoy seti, poluchennoy
Syndrome2_1 = Sigmoida2 (0,10 * Syndrome1_1-0,33 * Syndrome1_3)
Syndrome2_2 = Sigmoida2 (-0,22 * Syndrome1_1 + 0,36 * Syndrome1_2)
Syndrome 3-level:
X11 = (X11-4.5) / 4.5
X6
X22 = (X22-36.3) / 36.3
ELEC
CRANE
SVAR
FR X9
X10
X11 X22
Polya bazy dannyx (konechnye syndrome):
X10 = (X10-4.55) / 4.55
ELEC = (ELEC-0.5) / 0.5
Sigmoida1 (A) = A / (0,1+ | A |)
Sigmoida2 (A) = A / (0,1+ | A |)
Sigmoida3 (A) = A / (0,1+ | A |)
Syndrome 1 -go urovnya:
Functional preobrazovateli:
108
Machine Translated by Google

---

parameter. Further presented the structure of the neural network, separately for
vxodnÿe parameters podayutsya na vxod kajdogo neyrona v sloe); «Syndromes
izmeneniya temperatury sostavil, posle normirovaniya sostavlyaet [0, 1] (diapazon
mojet byt drugim v zavisimosti ot sposoba normirovaniya) - posle postobrabotki
temperatura mojet izmenyatsya ot 36 do 41 (ris. 15).
vsegda udobno dlya konechnogo polzovatelya, tak kak v bolshinstve zadach
signalami (znacheniyami vxodnyx poley) i 1-m sloem neyronov (kakie
nulya do edinitsy. Thus, the definition of "end syndromes" represents a significant
number in the specified range. It's not
X6 = 0.45 * Syndrom3_1 + 0.07 * Syndrom3_2-0.52 * Syndrom3_3-0.42
Post-processing of end-stage syndromes:
opredelennogo urovnya, kotorye ispolzuyutsya dlya vÿchisleniya vyxodnogo
izmeneniya. For example, trebuetsya obuchit prediktor dlya vÿchisleniya
2-go urovnya »- svyaz neyronov 1 sloya i 2 sloya; «Syndromes 3-go urovnya» - neurons
2 and 3 layers. The point "end syndromes" represents the sobo description of syndromes
Syndrome3_2 = Sigmoida3 (0.43 * Syndrome2_1)
Syndrome3_3 = Sigmoida3 (0.11 * Syndrome2_1)
Final syndromes:
trebuetsya poluchit znacheniya vyxodnogo polya so svoim diapazonom
interpreter answer sposoben otsenit uverennost neyronnoy seti v
poluchennom answer. Vector uverennosti seti in otvetax (for kajdogo otveta
opisÿvaet formula for normirovaniya znacheniy kajdogo vxodnogo
konechnomu polzovatelyu ispolzuetsya «interpretation answer» («postobrabotka»), t.
e. perevod «normirovannyx» znacheniy vyxodnogo polya
X6 = ((X6 * 33,70) +33,70) /
2) The first points of the verbal description ("field of data base") contain
perechen vxodnyx i vyxodnyx poley file dannyx. Predobrabotka
field. Sleduet uchityvat, chto pri ispolzovanii nekotoryx vidov otsenki
the patient's temperature, which in the norm is 36.6ÿ. Dlya vyvoda
kajdogo sloya: «syndromes 1-go urovnya» opisÿvayut svyazi mejdu vxodnymi
element vector uverennosti in otvete yavlyaetsya deystvitelnÿm chislom ot
in znacheniya iz isxodnogo diapazona (do predobrabotki dannyx). Range
svoya uverennost) mojet okazatsya poleznÿm dlya polzovatelya. Kajdyy
109
Machine Translated by Google

---

110
ne yavlyaetsya obyazatelnÿm. These operations serve to check the quality
range.
operatsiy i grupp operatsiy, traditsionno primenyaemyx issledovatelyami. Na
uchitelem dlya resheniya razlichnyx prikladnyx zadach. Described classification
T
normirovaniya, T - preobrazovanie otveta neyroimitatora iz diapazona [0, 1] (ili drugogo v
zavisimosti ot sposoba normirovaniya) T 'v isxodnyy
36
'
osnove opisaniya vxodnyx symptomov sloya i ix gruppirovki po neyronam
T
T
osnovnÿe operatsii i prosteyshie zadachi, realization kotoryx pozvolyaet
Polzovatel, posledovatelno proxodya sloy za sloem, mojet na
osnove abstraktnoy teknologii predlojena chastnaya metodika, prednaznachennaya prejde vsego
dlya nachinayushchix issledovateley. Privodyatsya
36
obyazatelnymi dlya vÿpolneniya. Strogo govorya, i perekrestnoe testirovanie
konkretnyx dannyx pri ispolzovanii gotovogo neyroimitatora.
davat nekotorÿe osmÿslennÿe imenovaniya i interpretatsii syndromov, generiruemym etim sloem.
Operatsii kontrastirovaniya i binarizatsii ne yavlyayutsya
41
dostatochno bÿstro i legko poluchit predvaritelnÿe rezultaty dlya
obucheniya setey (testirovanie) i uproshcheniya struktury seti
T - patient temperature (exhaust range), T '- posle
1
(contrast, binaryization). Output: Privedena
obobshchennaya technology
ispolzovaniya neyronnyx setey s
41
0
min
min
max
T ÿ T '(T ÿT ) ÿ T
Rice. 15. Scheme posledovatelnyx preobrazovaniy isxodnyx znacheniy
T
ÿ
TT
'ÿ
TT
ÿ
min
min
max
Machine Translated by Google

---

generation modeley opisÿvayushchix dannye v forme INS s uchitelem. Pri
dalee v tekste on budet upominatsya kak neyroserver. Neurogenesis may
2. Query processor; 3.
Database primerov; 4.
Base model;
iskopaemyx, prognoze povedeniya birj i dr. Vÿpolnyaetsya avtomaticheskaya
urovnya dekompozitsii:
1. Klient neyroservera;
Software complex Neurogenesis 1.0 (ROSPATENT testimony)
model parameters. Optimization structural INS osushchestvlyaetsya
6. Ispolnitel zadaniy;
generatsii program samostoyatelno opredelyaet optimalnuyu structure i